Multi-modal usability evaluation
Research into the usability of multi-modal systems has tended to be device-led, with a resulting lack of theory about multi-modal interaction and how it might differ from more conventional interaction. This is compounded by a confusion over the precise definition of modality between the various disciplines within the HCI community, how modalities can be effectively classified, and their usability properties. There is a consequent lack of appropriate methodologies and notations to model such interactions and assess the usability implications of these interfaces. The role of expertise and craft skill in using HCI techniques is also poorly understood. This thesis proposes a new definition of modality, and goes on to identify issues of importance to multi-modal usability, culminating in the development of a new methodology to support the identification of such usability issues. It additionally explores the role of expertise and craft skill in using usability modelling techniques to assess usability issues. By analysing the problems inherent in current definitions and approaches, as well as issues relevant to cognitive science, a clear understanding of both the requirements for a suitable definition of modality and the salient usability issues are obtained. A novel definition of modality, based on the three elements of sense, information form and temporal nature is proposed. Further, an associated taxonomy is produced, which categorises modalities within the sensory dimension as visual, acoustic and haptic. This taxonomy classifies modalities within the information form dimension as lexical, symbolic or concrete, and classifies the temporal form dimension modalities as discrete, continuous, or dynamic. This results in a twenty-seven cell taxonomy, with each cell representing one taxon, indicating one particular type of modality. This is a faceted classification system, with the modality named after the intersection of the categories, building the category names into a compound modality name. The issues surrounding modality are examined and refined into the concepts of modality types, properties and clashes. Modalities are identified as belonging to either the system or the user, and being expressive or receptive in type. Various properties are described based on issues of granularity and redundancy. The five different types of clashes are described. Problems relating to the modelling of multi-modal interaction are examined by means of a motivating case study based on a portion of an interface for a robotic arm. The effectiveness of five modelling techniques, STN, CW, CPM-GOMS, PUM and Z, in representing multi-modal issues are assessed. From this, and using the collated definition, taxonomy and theory, a new methodology, Evaluating Multi-modal Usability (EMU), is developed. This is applied to a previous case study of the robotic arm to assess its application and coverage. Both the definition and EMU are used by students in a case study to test the definition and methodology's effectiveness, and to examine the leverage such an approach may give. The results shows that modalities can be successfully identified within an interactive context, and that usability issues can be described. Empirical video data of the robotic arm in use is used to confirm the issues identified by the previous analyses, and to identify new issues. A rational re-analysis of the six approaches (STN, CW, CPM-GOMS, PUM, Z and EMU) is conducted in order to distinguish between issues identified through craft skill, based on general HCI expertise and familiarity with the problem, and issues identified due to the core of the method for each approach. This is to gain a realistic understanding of the validity of claims made by each method, and to identify how else issues might be identified, and the consequent implications. Craft skill is found to have a wider role than anticipated, and the importance of expertise in using such approaches emphasised. From the case study and the re-analyses the implications for EMU are examined, and suggestions made for future refinement. The main contributions of this thesis are the new definition, taxonomy and theory, which significantly contribute to the theoretical understanding of multi-modal usability, helping to resolve existing confusion in this area. The new methodology, EMU, is a useful technique for examining interfaces for multi-modal usability issues, although some refinement is required. The importance of craft skill in the identification of usability issues has been explicitly explored, with implications for future work on usability modelling and the training of practitioners in such techniques.